A key statistical quirk is missing from General Election polls in the UK, and it could decide the outcome

The Guardian's front page today proclaims that the polls have turned in Ed Miliband's favour: three consecutive polls have Labour marginally ahead of the Conservatives. Had they held the front page a little longer they would have had to update it - both a ComRes and YouGov poll yesterday evening showed precisely the opposite.

Due to the methodological differences between pollsters and the margin of error in any single poll, any claim that the latest figure marks a definitive shift should be treated with suspicion. Obviously, if a large number of polls showing strong gains for any of the parties it would shift the odds but, and I can't stress this enough, that is not what we're seeing.

In fact, what we're "not seeing" is the real news here: The polls are bucking a trend, the reversion to the mean, which statisticians usually expect to see appear at this time. And that is making the election much more complicated (and more exciting).

So what did we actually learn from these latest polls?

The best that we can say is that since the start of the year the Conservatives have eroded Labour's long-held lead and are now effectively neck-and-neck with their left-wing rivals.

That has not changed in weeks, and that may be the real news here.

Elsewhere, the Liberal Democrats appear to have stopped hemorrhaging support and have settled around 9% - although stronger local polling may see them hold more seats than the national swing against them would suggest.

Oh, and the SNP surge in Scotland looks very, very real. The UK looks almost certain to be heading for (at least) four-party politics.

Yet, despite the relative lack of recent movement in the polls, the expected seat count of each of the parties is moving - and this is hurting the two largest parties in particular, to the benefit of the SNP, largely.

The thing we are not seeing is the expected "reversion to the mean," the concept statisticians rely on to account for the fact that in the long-run, results even out over time.

Most forecasters include some amount of mean reversion into their models, based on what has happened in previous elections. This means that poll gains for non-incumbents tend to fade as the vote gets closer and incumbents gain accordingly.

In theory.

However, the closer we get to the date of the vote the more forecasters would expect to start seeing that mean reversion showing up in the polls. The longer the polls stay relatively stable the less likely it becomes that the assumed mean reversion will materialise. And with less than a month to go before May 7 stable polls are costing the Conservatives seats in England and Labour seats in Scotland.

That's big news in itself. The Conservatives need status quo bias to play a significant role in people's decision making if they are going to get back into government, and Labour need it too if they are to avoid a rout by the SNP north of Hadrian's wall.

Sadly, theory is not coming to the rescue of either party at this stage. And that alone might, just might, be decide the outcome.